Re: Why LL(1) Parsers do not support left recursion?

>As already mentioned with regards to the Java grammar, I think that a
>grammar with unrolled implied rules is a maintenance nightmare. How
>should any new kind of statement be added correctly, if the reason and
>the criteria for the splitting into open and closed statements is not
>documented explicitly, but instead is only built into an existing grammar?


This grammar solution is just an exercise of theoretical
interest. Yacc and most other tools can handle the ambiguous grammar
correctly.


>That's why I would prefer an explicit syntactical termination of an
>if-statement, by e.g. an "endif" or "fi" terminal, over your solution.
>This would IMO eliminate the dangling else problem, without the risk of
>collissions with other rules, in the current or any future extended
>grammar.


There is no need to modify the ambiguous grammar in any way to get a
correct parser. All that is needed is to choose one or the other of
the two possible parses. Yacc does this by preferring the shift over
the reduce. SLK does this by using the first production of the two
alternates as the parse table entry.

Return to the comp.compilers page.
Search the comp.compilers archives again.